duolahypercho

duolahypercho / andrej-karpathy-skills

Public

Codex-first conversion of Karpathy-inspired coding agent guidelines.

33
1
85% credibility
Found May 25, 2026 at 33 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
AI Summary

This project is a set of thoughtful guidelines that improve how AI coding assistants work. It takes advice from Andrej Karpathy about how AI should approach coding tasks—thinking carefully before diving in, keeping solutions simple, making only the changes needed, and verifying the work is actually done. The project packages these guidelines in two ways: as a plugin that connects directly to your AI tool, or as a simple copy-paste block you can add to your settings. Either way, the result is a more careful, reliable AI assistant that doesn't overbuild or make hidden assumptions.

How It Works

1
🤖 You use an AI coding assistant

You're working on code and using an AI helper to assist you, but sometimes it makes too many assumptions or overcomplicates things.

2
💡 You discover better guidelines

You find a collection of smart principles that help AI coding assistants think more carefully before making changes.

3
✨ You unlock thoughtful AI assistance

The guidelines teach the AI to pause, consider alternatives, keep things simple, and verify it's actually solved your problem.

4
Choose how to add the guidelines
🔌
Add as a plugin

Connect the guidelines package to your AI assistant so it reads them automatically whenever you work.

📋
Copy and paste instructions

Copy a special block of text and paste it into your AI assistant's settings for instant improvement.

5
🎯 Your AI thinks before acting

Now when you ask for help, the AI surfaces its assumptions, considers tradeoffs, and plans its approach first.

🎉 Better code, fewer surprises

Your AI assistant now makes careful, surgical changes that actually solve your problem without introducing unexpected issues.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 33 to 33 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is andrej-karpathy-skills?

This is a Codex plugin that packages Andrej Karpathy's coding agent guidelines into a reusable skill. It gives AI coding assistants four simple rules: think before coding, keep solutions minimal, make surgical changes only where needed, and define clear success criteria before finishing. The project ships as a Codex plugin you can install locally, or as a standalone Custom Instructions block you paste directly into settings. It includes examples and bilingual documentation.

Why is it gaining traction?

Developers are frustrated with AI coding assistants that overbuild, make unwarranted assumptions, and call work done without verification. This project addresses those exact pain points with a lightweight, opinionated framework. The fork from the original multica-ai repo adds Codex-native packaging, making it dead simple to install or copy-paste into your workflow. The dual delivery method (plugin or Custom Instructions) means you can use it without touching any repo files if you prefer.

Who should use this?

Developers using Codex who want more disciplined AI assistance. If your coding agent tends to overengineer solutions or edits code outside the scope of your request, these guidelines will help. Teams establishing coding agent standards will find the four principles useful as team-wide defaults. It's particularly valuable for code review workflows where surgical precision matters more than comprehensive rewrites.

Verdict

A focused, practical tool for taming overeager AI assistants. The 0.85% credibility score reflects early-stage maturity with modest community adoption at 33 stars, but the concept is solid and the implementation is straightforward. Worth trying if you use Codex and want more predictable, careful AI coding behavior.

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.